|Publication number||US7039696 B1|
|Application number||US 09/703,962|
|Publication date||2 May 2006|
|Filing date||31 Oct 2000|
|Priority date||31 Oct 2000|
|Publication number||09703962, 703962, US 7039696 B1, US 7039696B1, US-B1-7039696, US7039696 B1, US7039696B1|
|Inventors||Eric A. Pulsipher, Joseph R. Hunt|
|Original Assignee||Hewlett-Packard Development Company, L.P.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (13), Non-Patent Citations (1), Referenced by (1), Classifications (11), Legal Events (6)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present invention relates generally to computer networks. More particularly, it relates to a method and system for processing data used to describe network connectivity.
As communications networks, such as the Internet, carry more and more traffic, efficient use of the bandwidth available in the network becomes more and more important. Switching technology was developed in order to reduce congestion and associated competition for the available bandwidth. Switching technology works by restricting traffic. Instead of broadcasting a given data packet to all parts of the network, switches are used to control data flow such that the data packet is sent only along those network segments necessary to deliver it to the target node. The smaller volume of traffic on any given segment results in few packet collisions on that segment and, thus, the smoother and faster delivery of data. A choice between alternative paths is usually possible and is typically made based upon current traffic patterns.
The intelligent routing of data packets with resultant reduction in network congestion can only be effected if the network topology is known. The topology of a network is a description of the network which includes the location of and interconnections between nodes on the network. The word “topology” refers to either the physical or logical layout of the network, including devices, and their connections in relationship to one another. Information necessary to create the topology layout can be derived from tables stored in network devices such as hubs, bridges, and switches. The information in these tables is in a constant state of flux as new entries are being added and old entries time out. Many times there simply is not enough information to determine where to place a particular device.
Switches examine each data packet that they receive, read the source addresses, and log those addresses into tables along with the switch ports on which the packets were received. If a packet is received with a target address without an entry in the switches table, the switch receiving it broadcasts that packet to each of its ports. When the switch receives a reply, it will have identified where the new node lies.
In a large network with multiple possible paths from the switch to the target node, this table can become quite large and may require a significant amount of the switch's resources to develop and maintain. As an additional complication, the physical layout of devices and their connections are typically in a state of constant change. Devices are continually being removed from, added to, and moved to new physical locations on the network. To be effectively managed, the topology of a network must be accurately and efficiently ascertained, as well as maintained.
Existing mapping methods have limitations that prevent them from accurately mapping topological relationships. Multiple connectivity problems are one sort of difficulty encountered by existing methods. For example, connectors such as routers, switches, and bridges may be interconnected devices in a network. Some existing methods assume that these devices have only a single connection between them. In newer devices, however, it is common for manufacturers to provide multiple connections between devices to improve network efficiency and to increase capacity of links between the devices. The multiple connectivity allows the devices to maintain connection, in case one connection fails. Methods that do not consider multiple connectivity do not present a complete and accurate topological map of the network.
Another limitation of existing topology methods is the use of a single reference to identify a device. Existing methods use a reference interface or a reference address in a set of devices to orient all other devices in the same area. These methods assumed that every working device would be able to identify, or “hear,” this reference and identify it with a particular port of the device. With newer devices, however, it is possible that the same address or reference may be heard out of multiple ports of the same device. It is also possible that the address or reference may not be heard from any ports, for example, if switching technology is used.
Still another limitation of existing mapping systems is that they require a complete copy of the topological database to be stored in memory. In larger networks, the database is so large that this really is not feasible, because it requires the computer to be very large and expensive.
Still another difficulty with existing systems is that they focus on the minutia without considering the larger mapping considerations. Whenever an individual change in the system is detected, existing methods immediately act on that change, rather than taking a broader view of the change in the context of other system changes. For example, a device may be removed from the network temporarily and replaced with its ports reversed. In existing systems, this swapped port scenario could require hundreds or thousands of changes because the reference addresses will have changed for all interconnected devices.
Still another disadvantage of existing methods is that they use a continuous polling paradigm. These methods continuously poll network addresses throughout the day and make decisions based on those continuous polling results. This creates traffic on the network that slows other processes.
Still another limitation of existing methods is the assumption that network parts of a particular layer would be physically separated from other parts. Network layer 1 may represent the physical cabling of the network, layer 2 may represent the device connectivity, and layer 3 may represent a higher level of abstraction, such as the groupings of devices into regions. Existing methods assume that all layer 3 region groupings are self-contained, running on the same unique physical networking. However, in an internet protocol (IP) network, multiple IP domains may co-exist on the same lower layer networking infrastructure. It has become common for a network to employ a virtual local area network (LAN) to improve security or to simplify network maintenance, for example. Using virtual LANs, a system may have any number of different IP domains sharing the same physical connectivity. As a result, existing methods create confusion with respect to topological mapping because networks with multiple IP addresses in different subnets for the infrastructure devices cannot be properly represented because they assume the physical separation of connectivity for separate IP domains. Still another limitation of existing methods is that they do not allow topological loops, such as port aggregation or trunking, and switch meshing.
A software method is disclosed for processing data pertaining to connections in a communications network, such as a computer network. The data may be used to map the topology of the network to aid network management. The system creates data structures referred to as tuples to store the relationships between network nodes. A connection calculator receives tuple information from a tuple manager and creates additional tuples based on this data. The connection calculator identifies singly-heard host links, from which it then builds tuples to represent the network infrastructure. To build the infrastructure, the method creates tuples for the singly-heard host links, and then creates tuples for conn-to-conn links based on existing tuples and on hints derived from singly-heard host links tuples, which hints are maintained as extra host links tuples. The method then attempts to disprove invalid conn-to-conn links tuples and attempts to resolve conflicts between inconsistent tuples. The method creates tuples for nodes involving shared media connections. If the connection calculator cannot create a tuple because there is insufficient information about a connection, it requests additional information from that node. After the tuples are created, the connection calculator consolidates those binary tuples involving shared media connections into n-ary tuples to represent the shared media connection. The refined tuples may then be used to identify changes in the network topology.
The system provides an improved method for creating topological maps of communication networks based. Connectivity information is retrieved from the network nodes and stored as “tuples” to track specifically the desired information necessary to map the topology. These light weight data structures may store the host identifier, interface index, and a port. From this tuple information, the topology may be determined. A tuple may be a binary element insofar as it has two parts representing the two nodes on either end of a network link or segment. A “tuco” refers to a tuple component, such as half of a binary tuple.
As used herein, a node is any electronic component, such as a connector or a host, or combination of electronic components with their interconnections. A connector is any network device other than a host, including a switching device. A switching device is one type of connector and refers to any device that controls the flow of messages on a network. Switching devices include, but are not limited to, any of the following devices: repeaters, hubs, routers, bridges, and switches.
As used herein, the term “tuple” refers to any collection of assorted data. Tuples may be used to track information about network topology by storing data from network nodes. In one use, tuples may include a host identifier, interface information, and a port specification for each node. The port specification (also described as the group/port) may include a group number and a port number, or just a port number, depending upon the manufacturer's specifications. A binary tuple may include this information about two nodes as a means of showing the connectivity between them, whether the nodes are connected directly or indirectly through other nodes. A “conn-to-conn” tuple refers to a tuple that has connectivity data about connector nodes. A “conn-to-host” tuple refers to a tuple that has connectivity data about a connector node and a host node. In one use, tuples may have data about more than two nodes; that is, they may be n-ary tuples, such as those used with respect to shared media connections described herein.
A “singly-heard host” (shh) refers to a host, such as a workstation, PC, terminal, printer, other device, etc., that is connected directly to a connector, such as a switching device. A singly-heard host link (shhl) refers to the link, also referred to as a segment, between a connector and an shh. A “multi-heard host” (mhh) refers to hosts that are heard by a connector on the same port that other hosts are heard. A multi-heard host link (mhhl) refers to the link between the connector and an mhh. A link generally refers to the connection between nodes. A segment is a link that may include a shared media connection.
For illustrative purposes, nodes in the figures described above and in subsequent figures are shown as individual electronic devices or ports on connectors. Also, in the figures the nodes are represented as terminals. However, they could also be workstations, personal computers, printers, scanners, or any other electronic device that can be connected to networks 110.
The first, third and fourth hosts 121, 123, 124 are singly-heard hosts connected to separate ports 131, 133, 134 of a common connector 140—the first connector 140. The fifth and sixth hosts 125, 126 are singly-heard hosts connected to the third and fourth connectors 142, 143. The seventh and eighth hosts 127, 128 are multi-heard hosts connected to the same port 139 of the fifth connector 144. The multi-heard hosts 127, 128 illustrate a shared media segment 180, also referred to as a bus 180.
The second, third, fourth, and fifth connectors 141, 142, 143, 144 are interconnected and illustrate a switch mesh 181. Each of the connectors in the switch mesh 181 is connected to each other, either directly or indirectly, to create a fully meshed connection. In the mesh, traffic may be dynamically routed to create an efficient flow.
After the data is gathered and the tuples are stored in the neighbor database 310, the connection calculator 320 processes the tuples to reduce them to relationships in the topology.
If the connector hears only the one host on the port—that is, if the host is a singly-heard host—then the connection calculator 320 determines 408 whether the host is heard singly by any other connectors. If no other connectors hear the host as a singly-heard host, then the tuple is classified as a singly-heard host link (shhl) tuple 412 and other tuples for this host are classified 414 as extra host links (ehl). Another tuple for this host may be, for example, an intermediate connector connected indirectly to a host. For example,
The first weeding process also attempts to identify conflicts. If other connectors hear the host as a singly-heard host, then a conflict arises and the tuple is classified 410 as a singly-heard conflict link (shcl) tuple to be resolved later. This conflict may arise, for example, if a host has been moved within the network, in which case the forwarding table data may no longer be valid. Certain connectors previously connected directly to the host may still indicate that the moved host is connected. When all tuples have been processed 402 to identify singly-heard host links, the first weeding phase 922 is complete.
After processing extra host links for singly-heard host links, the connection calculator 320 considers 430 each connector (referred to as conn1) in the tuples to determine the relationship between connectors. As illustrated in
For every conn1, the connection calculator 320 considers 432 every other connector (conn2) to determine whether a conn1-to-conn2 tuple exists. If conn1-to-conn2 does not exist, then the connection calculator 320 considers 436 every other conn-to-conn tuple containing conn2. The other connector on this tuple may be referred to as conn3. If conn2 hears conn3 on a unique port 438 and if conn1 also hears conn3 440, then the connection calculator 320 creates 442 a tuple for conn1-to-conn2 in the connector-to-connector links tuple list.
After processing all of the conn1 tuples, the connection calculator 320 processes 444 each conn1-to-conn2 links tuple to ensure that they have complete port data. For each incomplete tuple 446, the connection calculator 320 looks 448 for a different tuple involving conn1 in the extra host links tuples on a different port. If a different tuple is found 450, then the connection calculator 320 determines 452 whether conn2 also hears the host. If conn2 does hear the host, then the connection calculator 320 completes the missing port data for conn2. If conn2 does not also hear the host 452, then the connection calculator 320 continues looking 448 through different tuples involving conn1 in extra host links on different ports.
After attempting to complete the missing data in each of the conn-to-conn links tuples, the connection calculator 320 processes 456 each conn-to-conn links tuple. The purpose of this sub-phase is to attempt to disprove invalid conn-to-conn links. The connection calculator 320 considers 458 conn1 and conn2 of each conn-to-conn links tuple. Every other connector in conn-to-conn links may be referred to as testconn. For each testconn 460, the connection calculator 320 determines 462 whether the testconn hears conn1 and conn2 on different groups/ports. If testconn hears conn1 and conn2 on different ports, then the tuple is moved to extraconnlinks (ecl) 464. Otherwise, the connection calculator 320 continues processing 460 the remaining testconns.
For each multi-heard host links (mhhl) tuple, also referred to as multiHeardLinks (mhl) tuples (sometimes referred to as the search tuple) 484, conn1 and host1 are considered 486. For each extra host links tuple containing host1 488, conn2 is considered 490. If there is a tuple in conn-to-conn links for conn2 and conn1 492, and if there is a conn2-to-host1 tuple in extraHostLinks 494, and if the group/port for conn2 hearing conn1 and host1 is different 496, then the search tuple is moved 498 to extraHostLinks.
The connection calculator 330 process described above collects the tuple information from the tuple manager 300, builds tuples new tuples and removes redundant or unnecessary tuples to produce the new topology. This topology may have incomplete tuples possibly resulting from extraneous information that the connection calculator 330 could not disprove. To refine the new topology, the connection calculator 330 can request the tuple manager 300 to obtain additional information about particular nodes or it may also request a user to refine the topology by adding or removing tuples. Using the process of the connection calculator 330, tuples marked as non-essential may be removed from the new topology to save space and to simply the topology. The connection calculator 330 is not confused by multiple connectivity situations such as port aggregation 182 or switch meshing 181 as shown in
After processing each mhhl tuple 518, each singly-heard host links (shhl) tuple, also referred to as a singlyHeardLinks (shl) tuple, is considered 540. For each shhl tuple, the connector and host are considered 542. If there is no existing singlyHeardSegments (shs) tuple for the connector 544, then an shs tuple is created 546. The host tuco is then added to the shs 548.
If the connector node is not connected to a star segment 556 and is connected to a bus segment 566, the topology converter 340 determines 568 whether there is an existing mhs tuple for conn1. If there is not an existing mhs tuple for conn1, then a topo mhs tuple is created 570. A tuco is added 572 for the host to the mhs tuple.
If the connector node is not connected to either a star segment 556 or to a bus segment 566, then the topology converter knows that it is connected to another connector (conn2). If such a connector does not already have an existing connLinks tuple for conn1 and conn2 576, then a connLinks tuple is created 578. After processing the bus segment, star segment, and conn-to-conn segment, for each conniface 554, the topology converter 340 proceeds to the next node 550.
If the node is not in a star segment, then the topology converter 340 knows that it is in the bus segment. If there is not already an mhs tuple for the node, 594, then the topology converter 340 creates 596 an mhs tuple. The tuco for the node is then added 598 to the mhs tuple, and the topology converter proceeds to the next node 550.
The system then processes 620 each unmarked tuple in the newSHL tuples. Four cases are possible for the host of the newSHL tuples. The host of the newSHL can be found in the current singlyHeardLinks (curSHL) 622, the current multiHeardLinks (curMHL) 630, the current connLinks (curCL) 638, or the current UnheardOfLnks (curUOL) 642. If the host of a new HL tuple is found 622 in the current SinglyHeardLinks (curSHL) tuples, then the system determines 624 if there is a matching connector tuco between the newSHL tuples and the curSHL tuples. If there is a matching tuco, then the system changes 626 the host connection attribute. If there is not a matching tuco, then the host connection is moved 628 in the topology.
If the host is found in the curMHL tuples 630, then the system determines 632 whether there is a matching connector tuco between the newSHL tuples and the curSHL tuples. If there is a 1 matching connector, then the segment type of connection is changed 634. If there is not a matching connector, then the host connection is moved 636 in the topology. If the host is found in the curCL tuples 638, then the host is moved 640 into a star segment of the connector. If it is found in the curUOL 642, then the host is moved 644 into the star segment of the connector.
If the host is found in the curMHL tuples 656, then the system determines 658 whether there is a matching connector tuco in both the curMHL tuples and the newMHL tuples. If there is a matching connector tuco, then the host connection attribute is changed 660. If there is not a matching tuco, then the host connection is moved 662 in the topology. If the host is found in the curCL tuples 664, then the host is moved into a bus segment of a connector. If the host is found in the curUOL tuples 668, then the host connection is moved 670 in the topology.
Another part of the identify different tuples phase 938 is shown in blocks 686 and 688 of
An advantage of the system is that it may be schedulable. The system may map network topology continuously, as done by existing systems, or it may be scheduled to run only at certain intervals, as desired by the user. A further advantage of the system is that it is capable of processing multiple connections between the same devices and of processing connection meshes, because it tracks each nodal connection independently, without limitations on the types of connections that are permitted to exist.
Although the present invention has been described with respect to particular embodiments thereof, variations are possible. The present invention may be embodied in specific forms without departing from the essential spirit or attributes thereof. It is desired that the embodiments described herein be considered in all respects illustrative and not restrictive and that reference be made to the appended claims for determining the scope of the invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5297138 *||30 Apr 1991||22 Mar 1994||Hewlett-Packard Company||Determining physical topology across repeaters and bridges in a computer network|
|US5450408 *||30 Sep 1991||12 Sep 1995||Hewlett-Packard Company||Method of ascertaining topology features of a network|
|US5727157 *||19 Dec 1996||10 Mar 1998||Cabletron Systems, Inc.||Apparatus and method for determining a computer network topology|
|US6108702 *||2 Dec 1998||22 Aug 2000||Micromuse, Inc.||Method and apparatus for determining accurate topology features of a network|
|US6295558 *||21 Aug 1998||25 Sep 2001||Hewlett-Packard Company||Automatic status polling failover or devices in a distributed network management hierarchy|
|US6405248 *||6 Jul 2000||11 Jun 2002||Micromuse, Inc.||Method and apparatus for determining accurate topology features of a network|
|US6625657 *||25 Mar 1999||23 Sep 2003||Nortel Networks Limited||System for requesting missing network accounting records if there is a break in sequence numbers while the records are transmitting from a source device|
|US6639900 *||15 Dec 1999||28 Oct 2003||International Business Machines Corporation||Use of generic classifiers to determine physical topology in heterogeneous networking environments|
|US6643699 *||10 Mar 2000||4 Nov 2003||International Business Machines Corporation||Computerized method for computing the overall splitting cost of a data structure using blocking island contour map of a network-level graph|
|US6671724 *||21 Mar 2000||30 Dec 2003||Centrisoft Corporation||Software, systems and methods for managing a distributed network|
|US6697338 *||28 Oct 1999||24 Feb 2004||Lucent Technologies Inc.||Determination of physical topology of a communication network|
|GB2350034A||Title not available|
|JP2000076209A||Title not available|
|1||Search Report Under Section 17 dated May 20, 2002 issued on corresponding application in GB in connection with application No. GB 0125510.8.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US20040151121 *||31 Jan 2003||5 Aug 2004||Srikanth Natarajan||Method of determining a maximal mesh|
|U.S. Classification||709/223, 709/221, 709/220, 709/224|
|International Classification||G06F15/173, H04L12/24, H04Q3/00|
|Cooperative Classification||H04Q3/0062, H04L41/12|
|European Classification||H04L41/12, H04Q3/00D4|
|12 Mar 2001||AS||Assignment|
Owner name: HEWLETT-PACKARD COMPANY, COLORADO
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:PULSIPHER, ERIC A.;HUNT, JOSEPH R.;REEL/FRAME:011617/0589
Effective date: 20001031
|30 Sep 2003||AS||Assignment|
Owner name: HEWLETT-PACKARD DEVELOPMENT COMPANY L.P., TEXAS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HEWLETT-PACKARD COMPANY;REEL/FRAME:014061/0492
Effective date: 20030926
Owner name: HEWLETT-PACKARD DEVELOPMENT COMPANY L.P.,TEXAS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HEWLETT-PACKARD COMPANY;REEL/FRAME:014061/0492
Effective date: 20030926
|2 Nov 2009||FPAY||Fee payment|
Year of fee payment: 4
|13 Dec 2013||REMI||Maintenance fee reminder mailed|
|2 May 2014||LAPS||Lapse for failure to pay maintenance fees|
|24 Jun 2014||FP||Expired due to failure to pay maintenance fee|
Effective date: 20140502